摘要
合成孔径雷达(SAR)数据包含观测目标的形状、方向、位置等丰富的地面信息,在林分类型识别和参数提取具有广泛的应用潜力。以福建省将乐实验林场2013年C波段超精细全极化单视复数RADARSAT-2数据为数据源,运用随机森林的分类方法和多种极化目标分解结果,对将乐国有林场的地物类型识别能力进行实验,分析4种极化方式的干涉相干性与林分类型识别精度的关系。结果表明:特征筛选可以使分类精度提高,总体分类精度由83.71%提高到87.43%,Kappa系数由0.8100提高到0.8533;干涉相干性作为分类特征提高了分类精度,总体分类精度为91.43%,Kappa系数为0.9000。
Synthetic aperture radar(SAR) data contains abundant ground information such as the shape, direction and location of the observed target, which has a wide application potential in forest type identification and parameter extraction. With the C-band super fine full polarization single view complex radarsat-2 data of Jiangle Experimental Forest Farm in Fujian Province in 2013, the classification method of random forest and the decomposition results of multi polarization targets were used to test the recognition ability of ground object types in Jiangle State-owned Forest Farm, and the relationship between interference coherence of four polarization methods and stand recognition accuracy was analyzed. The feature selection can improve the classification accuracy, the overall classification accuracy is increased from 83.71% to 87.43%, and the Kappa coefficient is increased from 0.810 0 to 0.853 3;the interference coherence as a classification feature significantly improves the classification accuracy, the overall classification accuracy is 91.43%, and the Kappa coefficient is 0.900 0.
作者
侯敬怡
张延成
范文义
Hou Jingyi;Zhang Yancheng;Fan Wenyi(Northeast Forestry University,Harbin 150040,P.R.China;Key Laboratory of Sustainable Forest Ecosystem Management-Ministry of Education,Northeast Forestry University)
出处
《东北林业大学学报》
CAS
CSCD
北大核心
2020年第11期33-38,共6页
Journal of Northeast Forestry University
基金
国家自然科学基金项目(31971654)
民用航天技术预先研究项目(D040114)
中央高校基本科研业务费专项资金项目(2572019CP12)。
关键词
全极化SAR
目标分解
干涉相干性
随机森林
Fully polarized SAR
Target decomposition
Interference features
Random forest
